Using the Mesoscale Model Evaluation Testbed (MMET) to test physic options in the Weather Research & Forecasting (WRF) model

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Sunday, 4 January 2015
Anthony D. Torres, University of Michigan/SOARS (UCAR), Ypsilanti, MI; and J. K. Wolff, M. Harrold, C. L. Phillips, and D. J. Posselt

The Mesoscale Model Evaluation Testbed has been set up by the Developmental Testbed Center (DTC) to assist the research community in efficiently testing and evaluating newly developed model techniques aimed to more accurately predict the weather and to potentially be implemented into operations. For this project, datasets available through the MMET were utilized to test the forecast performance of several configurations of the Weather Research & Forecasting model (WRF v3.5.1) using the Advanced Research WRF dynamical core for different physical parameterization schemes and grid-spacing. A significant derecho event that occurred on 29 June 2012 over the U.S. Midwest and Mid-Atlantic states was chosen for this case study. Statistical analysis of each WRF configuration was conducted using the Model Evaluation Tools (MET) and results were plotted using R, a statistical package. Standard verification metrics were calculated for surface and upper-air temperature, dew point temperature, and wind speed, as well as precipitation. In addition, a more advanced spatial verification technique, known as the Method for Object-based Diagnostic Evaluation (MODE), was utilized to diagnose errors in forecast precipitation placement, coverage, and orientation. Results indicated that model performance was sensitive to grid-spacing and physics options. Adjusting the grid-spacing from 15- to 5-km, while utilizing the same physics options, allowed for strong convective development that otherwise did not occur. Substituting different microphysics and radiation options at the 5-km grid-spacing had significant effects on developing the storm and resulting storm structure, while changing the planetary boundary layer scheme positively impacted the storm placement. The synoptic environment depicted by the WRF configurations that were able to simulate strong convective development is consistent with research on derecho formation. Figure 1. Example illustrating MODE objects created from WRF 3-h accumulated precipitation fields and the associated Stage II analysis fields for the 39 hour forecast from the 28 June 2012 12 UTC initialization. In this example, the configuration being evaluated is using the RRTMG long- and short-wave radiation schemes. A convolution threshold of 0.254 mm was used to define the objects; the shaded regions represent objects in the forecast field, and outlined areas represent objects in the observed field.